The IBM 2015 English Conversational Telephone Speech Recognition System
This work addresses speech recognition accuracy for telephone conversations, representing an incremental improvement in a domain-specific application.
The paper tackles improving English conversational telephone speech recognition by applying techniques like maxout networks with annealed dropout, large-scale training, joint modeling of neural networks, and advanced language model rescoring, resulting in an 8.0% word error rate on the Switchboard test set, a 23% relative improvement over previous results.
We describe the latest improvements to the IBM English conversational telephone speech recognition system. Some of the techniques that were found beneficial are: maxout networks with annealed dropout rates; networks with a very large number of outputs trained on 2000 hours of data; joint modeling of partially unfolded recurrent neural networks and convolutional nets by combining the bottleneck and output layers and retraining the resulting model; and lastly, sophisticated language model rescoring with exponential and neural network LMs. These techniques result in an 8.0% word error rate on the Switchboard part of the Hub5-2000 evaluation test set which is 23% relative better than our previous best published result.